binary_model_cluster: Wrapper for a binary executable for parallel simulations

In EasyABC: Efficient Approximate Bayesian Computation Sampling Schemes

Wrapper for a binary executable for parallel simulations

Description

This function enables to link a binary executable to a R function.

Usage

  binary_model_cluster(command)

Arguments

command

a character string indicating the command to launch the executable on your system.

Details

A binary executable used with binary_model_cluster has to be placed in the current directory of the R session. It further has to respect several constraints: 1- it has to have a single argument: a number k used by the binary executable to know in which files read and write. 2- it has to read the seed for its pseudo-random number generator and the model parameters in a file "inputk" (where k is the argument passed to the executable: input1, input2,...). 3- it has to write the summary statistic in a file "outputk" (where k is the argument passed to the executable: output1, output2,...). The file "inputk" will be generated by the wrapper binary_model_cluster in the current directory of the R session, and the wrapper will read the file "outputk" generated by the binary executable. This construction ensures that each core reads and writes in different files. In the file "inputk", the first line contains the seed, and each subsequent line contains one model parameter value. In the file "outputk", each summary statistic should be separated by a space or a tab separation. This wrapper should be used for use with multiple cores of the computer. If the user wishes to use a single core of the computer, the wrapper binary_model should be used. Note that the files "inputk" and "outputk" are deleted by the wrapper at the end of the function.

Value

A R function wrapping the binary executable, to be used with the EasyABC functions.

Author(s)

Franck Jabot, Thierry Faure and Nicolas Dumoulin

See Also

binary_model, ABC_rejection, ABC_sequential, ABC_mcmc

Examples

 
 ## Not run: 
    ## artificial example to show how to use the binary_model function with
    # an executable "My_Executable"
    ABC_rej<-ABC_rejection(model=binary_model_cluster("./My_Executable"),
      prior=..., n_cluster=2,...)
    # NB: on windows, "My_Executable" should be of the form "My_Executable.exe" :
    ABC_rej<-ABC_rejection(model=binary_model_cluster("./My_Executable.exe"),
      prior=..., n_cluster=2,...)
 
## End(Not run)

EasyABC documentation built on Jan. 6, 2023, 1:14 a.m.